# Change: Add Omni Intelligence Hub

## Why
TrueGrowth needs an upstream intelligence workspace that helps users decide what to create before they generate images, videos, voices, avatars, or canvas workflows. The provided product plan positions "万象智枢 / Omni Intelligence Hub" as the platform's global research and content strategy center, covering web, video, community, repository, trend, competitor, and local document intelligence.

Agent-Reach already targets the required retrieval layer: web pages, YouTube, RSS, search, GitHub, and authenticated community sources. Integrating it as a local runtime bridge lets TrueGrowth add this intelligence entry without duplicating crawler and source acquisition logic in the workbench.

## What Changes
- Add a left-sidebar entry and route for "万象智枢".
- Add an application-center card for the new module.
- Add an Omni Intelligence Hub page with the planned navigation groups: 总览, 探索, 洞察, 创作, 管理.
- Add a unified query surface with modes for 快速探索, 深度研究, 趋势分析, 竞品分析, and 链接解析.
- Reuse the current canvas/text model selection logic so users can choose the analysis model from runtime-discovered text models.
- Add an Agent-Reach runtime bridge for capability checks, diagnostics, and research/link-analysis task submission.
- Show a usable degraded state when Agent-Reach is not installed, not authenticated, or missing source capabilities.
- Preserve results as structured intelligence assets that can later feed scripts, storyboards, prompts, digital-human copy, video plans, canvas insertion, and the asset center.

## Impact
- Affected specs: omni-intelligence-hub
- Affected code:
  - `apps/web/src/workbench/WorkbenchShell.tsx`
  - `apps/web/src/workbench/app-definitions.tsx`
  - `apps/web/src/workbench/workbench.scss`
  - `apps/web/src/workbench/truegrowth.types.ts`
  - local runtime scripts and bridge endpoints under `scripts/` or `apps/web/src/workbench` service helpers
  - reusable model selector imports from `packages/drawnix/src/components/chat-drawer` and `packages/drawnix/src/hooks/use-runtime-models`